Disclosure
Every provider. Every model. Every feature. This is the complete and real list of what runs behind the scenes on EL7.AI, taken directly from production configuration — not a marketing brochure.
Last updated: April 26, 2026
EL7.AI runs 6 AI providers and 13 distinct models today, serving over 90 separate tasks across the platform. Every task is routed through the internal EL7 LLM Router, which picks the right model based on task type, cost, and latency. This page lists the full set, because transparency is a precondition for trust.
All providers below are actively used in production. We contract only with providers offering contractual commitments not to retain queries and not to use them to train their models.
Provider of Claude. Used for deep long-context analysis and precise editorial writing tasks.
Provider of the Gemini family. Our most-used provider (~64% of tasks), thanks to a strong balance of quality, latency, and cost — especially for Arabic translation and summarization.
Provider of the GPT family. Used for structured extraction, classification, headline generation, and conversational assistant features.
Provider of Grok. Used for news analysis and fast logical reasoning tasks.
Used for web search with source citations, particularly for refreshing information about recent events.
Backup for math/logic tasks when needed. Used in narrow circumstances and never on personal data.
This is an accurate list of the models actually used in production task configurations (snapshot from llm_task_configs).
| Model | Provider | Tasks | Where it's used |
|---|---|---|---|
| Gemini 3 Flash Preview | 64 tasks | Translation, news summaries, event analysis, weekly COT report cards | |
| Gemini 3 Pro Preview | 37 tasks | Deep Fed data analysis, Beige Book analysis, editorial article writing | |
| GPT-5.4 (2026-03-05) | OpenAI | 17 tasks | Pre-FOMC analysis, entity extraction, economic intelligence reports |
| GPT-4o-mini | OpenAI | 10 tasks | Light classification tasks, conversation titles, keyword variants |
| GPT-5.4 | OpenAI | 10 tasks | Smart conversational assistant, SEO optimization, fast summaries |
| Grok 4 Fast Reasoning | xAI | 3 tasks | Deep text analysis, interpreting unusual market moves |
| Sonar Pro | Perplexity | 2 tasks | Cited web search to refresh news and recent events |
| GPT-4o | OpenAI | 3 tasks | Fallback when GPT-5.4 is degraded |
| Gemini 2.0 Flash | 3 tasks | Low-latency fallback tasks | |
| Gemini 3.1 Flash Lite Preview | 2 tasks | Fast structured extraction tasks |
Every machine-generated piece of content carries a tag. We never hide AI use.
We do not share your data with LLM vendors for training. Queries pass through in zero-retention mode where available.
All outputs are for educational purposes. Financial decisions remain the user’s responsibility alone.
Educational content and courses are human-reviewed before publication. News summaries and analyses publish automatically with a clear tag.
If you spot inaccurate content or have a suggestion, contact us.